[Ffmpeg-devel-irc] ffmpeg-devel.log.20160507

burek burek021 at gmail.com
Sun May 8 02:05:03 CEST 2016


[05:12:30 CEST] <cone-267> ffmpeg 03Michael Niedermayer 07master:cbe265ccd714: avcodec/m101: Simplify if() condition
[05:12:30 CEST] <cone-267> ffmpeg 03Michael Niedermayer 07master:22d1148c77c6: avcodec/m101: remove unneeded zeroing of priv_data_size
[05:12:30 CEST] <cone-267> ffmpeg 03Michael Niedermayer 07master:fc6894770ea9: avcodec/m101: simplify 8bit code
[05:12:30 CEST] <cone-267> ffmpeg 03Piotr Bandurski 07master:e601e078e94f: avformat/riff: add M102 FourCC
[09:50:24 CEST] <Easyfab> delogo with  x=0 or  y=0 crash ffmpeg for me, can someone reproduce ?
[10:14:45 CEST] <durandal_170> what version?
[10:18:13 CEST] <Easyfab> http://pastebin.com/vK45biQs
[12:27:23 CEST] <nevcairiel> Daemon404: all the extra commits i send to the ML have been approved by michael, so if you want feel free to merge my branch (its even rebased on top of current master right now)
[12:37:32 CEST] <Compn> fatal: remote error: access denied or repository not exported: /ffmpeg.git
[12:37:44 CEST] <Compn> hmmm
[12:37:51 CEST] <Compn> screwed up my repo again
[12:41:07 CEST] <Compn> and the git instructions are not on the developer documentation page :P
[12:45:57 CEST] <Compn> $ git remote -v
[12:45:57 CEST] <Compn> origin  git://source.ffmpeg.org/ffmpeg.git (fetch)
[12:45:57 CEST] <Compn> origin  git://source.ffmpeg.org/ffmpeg.git (push)
[12:50:12 CEST] <wm4> went to the website, did 1 click, saw git clone https://git.ffmpeg.org/ffmpeg.git ffmpeg
[12:52:03 CEST] <Compn> yeah i remember it changed :P
[12:52:09 CEST] <Compn> but thats about it 
[12:53:07 CEST] <Compn> wm4 : also , i'm finding this page needs to be updated. https://ffmpeg.org/git-howto.html#Cloning-the-source-tree
[12:53:26 CEST] <Compn> but i guess thats my job
[12:54:16 CEST] <wm4> source still works, I'm using it
[12:58:06 CEST] <cone-422> ffmpeg 03compn 07master:1e109bbbf133: riff: add comment for g721,g723 codec tags
[13:00:04 CEST] <Compn> ah missed "avformat"
[13:00:21 CEST] <Compn> 1l coffee
[13:02:36 CEST] <Compn> wm4 : michaelni has provided proper git push url.
[13:07:40 CEST] <michaelni> Daemon404, nevcairiel the branch seems to break: http://samples.ffmpeg.org/V-codecs/smv2/ref_10.avi
[13:08:13 CEST] <michaelni> [h264 @ 0x7f5fc40034a0] No start code is found. ...[h264 @ 0x7f5fc40034a0] Error splitting the input into NAL units.
[13:08:25 CEST] <michaelni> with ./ffplay ~/videos/sm2v/ref_10.avi
[13:08:45 CEST] <JEEB> hmm, also nal unit related stuff reminds me of the fix required for the avc bsf
[13:09:01 CEST] <JEEB> I don't think the guy posted a new version after TEP2
[13:09:30 CEST] <JEEB> https://github.com/VFR-maniac/ffmpeg/commit/172fe56b1a4a063ff2feb14d3394129eb155b40d
[13:17:29 CEST] <nevcairiel> michaelni: it looks like it fails extradata parsing (both before and after), except that this is now considered a fatal error, and i'm not sure at which  point it should ignore th error for extradata
[13:19:26 CEST] <nevcairiel> perhaps ff_h264_decode_extradata should not check the return value of decode_nal_units
[13:21:10 CEST] <nevcairiel> previously it would just silently skip everything if no NALs could be extracted
[13:23:05 CEST] <nevcairiel> (removing the error condition makes the file works as expected, as only the extradata is broken)
[13:23:33 CEST] <nevcairiel> so, suggestions?
[13:30:30 CEST] <nevcairiel> michaelni: https://github.com/Nevcairiel/FFmpeg/commit/5f942f43d94e2be1e2653aa311f6e3fd2e303054 like this maybe
[13:33:43 CEST] <michaelni> nevcairiel, is it intended that this produces error messages where there where none before ?
[13:34:01 CEST] <nevcairiel> probably
[13:34:04 CEST] <nevcairiel> it is technically invalid data
[13:34:47 CEST] <nevcairiel> the old h264 code just silently discarded any garbage data it found
[13:36:04 CEST] <nevcairiel> the new h264/5 code errors out when a buffer is garbage-only, ie. not one valid NAL
[13:36:53 CEST] <nevcairiel> the only difference is error handling here, not actual parsing changes from what i can tell
[13:37:19 CEST] <michaelni> the codec tag says SMV2, so iam not sure that is "garbage" for SMV2
[13:37:33 CEST] <michaelni> i dont know if that is intended to be actual spec compliant h264
[13:37:44 CEST] <nevcairiel> well its garbage for the h264 decoder either way
[13:37:45 CEST] <nevcairiel> :)
[13:37:52 CEST] <michaelni> yes
[13:38:21 CEST] <michaelni> no objections but id like to do something about the errors at some point
[13:39:15 CEST] <michaelni> we also donz display errors for msmpeg4 not being mpeg4 syntax
[13:39:40 CEST] <michaelni> but i dont know what SMV2 is maybe its intended to be h264 exactly
[13:40:28 CEST] <nevcairiel> i think its generally a good idea to check the nal parsing a bit more thoroughly then before, but adding cases where the checks are silent would require adding parameters to all the involved functions
[13:41:55 CEST] <nevcairiel> anyway i'm going out into the sun for a bit, be back later
[13:42:12 CEST] <michaelni> sure, have fun!, ... if AVCodecContext is passed, AVCodecContext->codec_tag could be used for such checks maybe
[14:06:33 CEST] <Daemon404> ill meege tomight
[14:06:38 CEST] <Daemon404> im also out in the sun
[14:06:50 CEST] <Daemon404> in fact sitting in a park with a beer.
[14:07:16 CEST] <Daemon404> (i realie this sounds idd to americans)
[14:08:21 CEST] <wm4> Daemon404: how so
[14:09:03 CEST] <Daemon404> it is illegal to drink outside in .ca and .us
[14:09:19 CEST] <Daemon404> in public anyway
[14:09:24 CEST] <wm4> even beer?
[14:09:30 CEST] <Daemon404> yes
[14:09:37 CEST] <Daemon404> drink as in alcohol
[14:09:53 CEST] <Daemon404> im half way in between
[14:09:57 CEST] <Daemon404> i kike a beer in the sun
[14:10:09 CEST] <Daemon404> but i also see how ridiculous brits get outside with beer
[14:10:13 CEST] <Daemon404> so..
[14:10:17 CEST] <Daemon404> :V
[14:12:14 CEST] <JEEB> cold beer in the sun <3
[14:13:16 CEST] <Daemon404> there is sun in finland?
[14:13:16 CEST] <Compn> Daemon404 : is your beer in brown paper bag? 
[14:13:25 CEST] <Daemon404> no lol
[14:13:27 CEST] <Compn> thats how many people drink, with paper covering the beer bottle/can 
[14:13:29 CEST] <Compn> in usa
[14:13:33 CEST] <Daemon404> im not a hobo
[14:13:47 CEST] <Compn> well then park ranger will probably give you a few tickets :P
[14:16:42 CEST] <JEEB> Daemon404: towards summer we're having lots of sun but not necessarily heat
[14:16:54 CEST] <JEEB> heat being at around +20 here right now is kind of exceptional
[14:17:39 CEST] <Daemon404> 20 here too
[14:17:43 CEST] <Daemon404> its perfect weather
[14:17:49 CEST] <Daemon404> too bad summer will be 30
[14:17:49 CEST] <JEEB> yeah
[14:19:06 CEST] <Compn> get out and play some disc golf :)
[14:32:21 CEST] <Daemon404> disc golf? is that anything like ultimate frisbee?
[14:32:34 CEST] <Daemon404> well i guess it cant be
[14:33:05 CEST] <JEEB> there's usually a basketball bag kind of thing somewhere and then you have a starting point
[14:33:20 CEST] <JEEB> we did it at my previous job at a random all-office meet-out
[14:33:32 CEST] <JEEB> and you have a "par" etc
[14:33:48 CEST] <JEEB> (how many throws until bag is the number)
[14:35:56 CEST] <Shiz> can't wait for summer
[14:36:07 CEST] <wm4> can't wait for winter
[14:38:32 CEST] <Daemon404> ^
[14:46:37 CEST] <michaelni> Daemon404, nevcairiel if you merge take teh 2 commits or equivalent/better fixes from HEAD of https://github.com/michaelni/FFmpeg/commits/h264merge
[14:47:41 CEST] <nevcairiel> maybe the parser should be changed to skip zero sized NALs
[14:47:48 CEST] <nevcairiel> makes no sense to have them
[14:48:03 CEST] <nevcairiel> even the content-less NALs should be one byte in size for the nal type etc
[14:48:16 CEST] <michaelni> sounds reasonable
[15:51:21 CEST] <Daemon404> michaelni / nevcairiel - so am i clear to merge, or should i wait for this 0-sized thing
[16:30:16 CEST] <nevcairiel> Daemon404: i just pushed https://github.com/Nevcairiel/FFmpeg/commit/336b453aa270c03bf2972ec6e603e66b5292c66e which should do that
[16:31:41 CEST] <nevcairiel> NALs should generally always be 1 byte for h264 or 2 for hevc, as thats the size of the NAL header
[16:31:44 CEST] <nevcairiel> so zero is just invalid
[16:36:28 CEST] <Daemon404> nevcairiel, all right
[16:36:31 CEST] <Daemon404> ill get to pushing soon
[16:36:34 CEST] <Daemon404> working on hobby project atm
[16:36:42 CEST] <Daemon404> you can push if you want
[17:39:49 CEST] <nevcairiel> BtbN: could you post your configuration details to the ML when you get the internal compiler error with O3? in the "configure: enable GCC vectorization" thread? some people are interested :)
[17:43:00 CEST] <BtbN> nevcairiel, I didn't do anything special. Just --enable-libx264 --enable-gpl, on Docker alpine:3.3
[17:43:07 CEST] <BtbN> Which is using gcc 5.3.0
[17:50:10 CEST] <jamrial> can you still reproduce it?
[17:50:45 CEST] <cone-749> ffmpeg 03Marton Balint 07master:215a2d76788f: ffplay: force setting alsa buffer size
[17:50:45 CEST] <cone-749> ffmpeg 03Jan Sebechlebsky 07master:a3c877aca78f: avformat/tee: Fix TeeSlave.bsfs pointer array size
[17:50:45 CEST] <cone-749> ffmpeg 03Jan Sebechlebsky 07master:1bc83f6ea8ff: avformat/tee: Use ref instead copy in write_packet
[17:53:15 CEST] <BtbN> jamrial, i switched the container to Debian to avoid it, but I'm quite sure it would still happen, even with a plain configure without any arguments.
[17:55:07 CEST] <BtbN> yeah, still happens
[17:55:19 CEST] <jamrial_> on debian?
[17:55:23 CEST] <BtbN> alpine
[17:55:27 CEST] <BtbN> Debian has 4.9, that's fine.
[17:55:43 CEST] <jamrial_> i can't reproduce it on archlinux, and neither does any of the many fate clients
[17:55:56 CEST] <BtbN> I'll push the broken docker image somewhere
[17:56:16 CEST] <jamrial_> can you add -freport-bug to the gcc command line options?
[17:56:31 CEST] <jamrial_> it will generate a pre-processed output in /tmp
[17:56:56 CEST] <BtbN> so just ./configure --extra-cflags=-freport-bug ?
[17:57:44 CEST] <jamrial_> that, or use make V=1 to get the command and add that option to run gcc manually
[17:57:51 CEST] <jamrial_> so you don't have to reconfigure
[17:58:00 CEST] <nevcairiel> he is on linux, reconfigure takes like 5 seconds
[17:58:08 CEST] <BtbN> It's a docker container, it's rebuilt entirely every time
[17:58:08 CEST] <BtbN> ls
[17:58:16 CEST] <jamrial_> ok
[18:06:00 CEST] <BtbN> jamrial_, pushed the affected image, if you want to test around in it: docker run -ti btbn/ffmpeg_compiler_panic /bin/bash
[18:07:13 CEST] <jamrial_> where? i'm not familiar with docker
[18:07:46 CEST] <BtbN> If you have Docker installed, just run that command and it gives you a shell, with the ffmpeg source tree in /root/ffmpeg/
[18:08:25 CEST] <BtbN> I'm trying to upload the log it generates in tmp, but getting it out of the container is surprisingly annoying
[18:11:07 CEST] <BtbN> jamrial_, https://btbn.de/files/crashlog.out
[18:11:11 CEST] <BtbN> nevcairiel, ^
[18:14:19 CEST] <jamrial_> BtbN: can't reproduce the crash with that output either. it compiles it just fine
[18:14:44 CEST] <BtbN> So probably something alpine did to their gcc?
[18:15:48 CEST] <jamrial_> could be
[18:16:33 CEST] <BtbN> https://bugs.alpinelinux.org/issues/5334 they don't seem to care too much
[18:17:00 CEST] <nevcairiel> their certificate also expired, are they still alive? =P
[18:18:06 CEST] <wm4> last release 2 months ago
[18:20:08 CEST] <JEEB> lol
[18:21:33 CEST] <nevcairiel> i dont even know what half of their gcc build options do, and they are not mentioned in the gnu docs
[18:21:57 CEST] <nevcairiel> are they one of those ricer distros with 100 patches on top of everything?
[18:23:07 CEST] <BtbN> the entire goal of that distro is to be as small as possible.
[18:23:15 CEST] <jamrial_> save for --disable-fixed-point the rest i think are also used by arch's gcc package
[18:23:45 CEST] <nevcairiel> they do use musl, which gcc only claims to officially support as a target since 6.1
[18:23:51 CEST] <nevcairiel> at least according to the changelog
[18:24:38 CEST] <jamrial_> then i wont bother submitting a bug report. can't even reproduce it and they will just tell me it's not supported anyway
[18:25:53 CEST] <nevcairiel> unfortunately that seems to be the general response from many projects, they just don't care
[20:23:54 CEST] <DSM_> durandal_170: hi
[20:24:20 CEST] <DSM_> durandal_170: https://github.com/dsmudhar/FFmpeg/commit/45112cdb8ea812ab2264d65c5c7b3ea5238ce6dd
[20:24:21 CEST] <durandal_170> hi
[20:24:28 CEST] <DSM_> seems good?
[20:25:07 CEST] <DSM_> i categorized the MV types and frame types. 
[20:26:30 CEST] <durandal_170> where are I frames?
[20:27:07 CEST] <DSM_> adding :P
[20:27:14 CEST] <DSM_> example: -vf codecview=mv=fp+bp:frames=pf+bf
[20:30:05 CEST] <DSM_> i-frame can have both direction vectors?
[20:30:13 CEST] <DSM_> durandal_170: ping
[20:51:34 CEST] <durandal_170> DSM_: yes 
[20:51:48 CEST] <DSM_> ok
[21:05:09 CEST] <DSM_> durandal_170: https://github.com/dsmudhar/FFmpeg/commit/b5c83eeee6af345a6dc2f44c7db5cb9c2800625b
[21:05:19 CEST] <DSM_> added i-frame option
[21:07:04 CEST] <durandal_170> send to ml
[21:09:14 CEST] <DSM_> sending
[21:56:17 CEST] <jamrial_> oh nice, archlinux updated to gcc 6.1
[22:02:22 CEST] <fritsch> so see you in 3 days? after recompling the whole system 3 times?
[22:08:03 CEST] <Daemon404> fritsch, arch is binary packages
[22:15:19 CEST] <Daemon404> nevcairiel, do i still need to add michael's commits, or do your latest ones fix that
[22:15:56 CEST] <Daemon404> ah yeah, no i dotn need them
[22:22:54 CEST] <cone-749> ffmpeg 03Anton Khirnov 07master:ca2f19b9cc37: h264: switch to h2645_parse for NAL parsing
[22:22:55 CEST] <cone-749> ffmpeg 03Hendrik Leppkes 07master:6eb1b40ad869: hevc: fix size condition in ptl parsing
[22:22:56 CEST] <cone-749> ffmpeg 03Hendrik Leppkes 07master:c80238939309: h2645_parse: initialize the GetBitContext to the proper size
[22:22:57 CEST] <cone-749> ffmpeg 03Hendrik Leppkes 07master:9cc1ab63ac0f: h2645_parse: allow partial escaping
[22:22:58 CEST] <cone-749> ffmpeg 03Hendrik Leppkes 07master:772ad7142dff: Merge commit 'ca2f19b9cc37be509d85f05c8f902860475905f8'
[22:22:59 CEST] <cone-749> ffmpeg 03Hendrik Leppkes 07master:d6f92103e0f6: h264: do not return an error when NAL parsing of extradata failed
[22:23:00 CEST] <cone-749> ffmpeg 03Hendrik Leppkes 07master:d46e85635070: h265_parse: skip zero sized NAL units
[22:47:46 CEST] <omerjerk> hi everyone.
[22:48:00 CEST] <omerjerk> I need to analyze the file created by my decoder.
[22:48:16 CEST] <omerjerk> by file, I mean the wav file.
[22:48:49 CEST] <omerjerk> By analyzing, I mean to find out what's the sampling format etc.
[22:49:05 CEST] <omerjerk> Any tips for a tool to do this ?
[22:51:03 CEST] <cone-749> ffmpeg 03Vittorio Giovara 07master:9f4d99138df4: fate: Add test for WMV2 with jframes
[22:51:05 CEST] <cone-749> ffmpeg 03Vittorio Giovara 07master:d0540fd02171: intrax8: Pass macroblock size to ff_intrax8_common_init
[22:51:05 CEST] <cone-749> ffmpeg 03Vittorio Giovara 07master:b1268e0f032a: intrax8: Pass macroblock coordinates to ff_intrax8_decode_picture
[22:51:06 CEST] <cone-749> ffmpeg 03Vittorio Giovara 07master:c2084ffcbfc1: intrax8: Use the generic horizband function
[22:51:07 CEST] <cone-749> ffmpeg 03Vittorio Giovara 07master:9fa888c02801: intrax8: Keep a reference to the decoder blocks
[22:51:08 CEST] <cone-749> ffmpeg 03Vittorio Giovara 07master:9b57995cdd48: intrax8: Drop MB emulation code
[22:51:09 CEST] <cone-749> ffmpeg 03Diego Biurrun 07master:6ebd06a9b250: intrax8: Drop lots of pointless parentheses
[22:51:10 CEST] <cone-749> ffmpeg 03Vittorio Giovara 07master:ca8c7591735c: intrax8: Remove mpegvideo dependency
[22:51:11 CEST] <cone-749> ffmpeg 03Derek Buitenhuis 07master:eee5a7cd7d15: Merge commit '9f4d99138df434a73b097c997fb4cafc65f4ff54'
[22:51:12 CEST] <cone-749> ffmpeg 03Derek Buitenhuis 07master:578fb5a27d45: Merge commit 'd0540fd02171a6233d2016b199d013299debf7e3'
[22:51:13 CEST] <cone-749> ffmpeg 03Derek Buitenhuis 07master:123fef54cc8f: Merge commit 'b1268e0f032a3af3912fe3fb8d3855e12d7ea83b'
[22:51:14 CEST] <cone-749> ffmpeg 03Derek Buitenhuis 07master:c59c7924707e: Merge commit 'c2084ffcbfc11d1b6ed3a4a0df9cafd56fbb896f'
[22:51:15 CEST] <cone-749> ffmpeg 03Derek Buitenhuis 07master:f110c624b1f3: Merge commit '9fa888c02801fff2e8817c24068f5296bbe60000'
[22:51:16 CEST] <cone-749> ffmpeg 03Derek Buitenhuis 07master:741960e2f330: Merge commit '9b57995cdd489a4cff51dcc1a1f08ac77ec5a58c'
[22:51:17 CEST] <cone-749> ffmpeg 03Derek Buitenhuis 07master:f2c78128395e: Merge commit '6ebd06a9b2508747a135ee4c880d8f612e08932b'
[22:51:18 CEST] <cone-749> ffmpeg 03Derek Buitenhuis 07master:e811ebcd9cd5: Merge commit 'ca8c7591735c0f80cc29e31e2e92cb10228e14c7'
[22:51:20 CEST] <Daemon404> omerjerk, ffprobe?
[23:04:40 CEST] <omerjerk> I haven't used ffmpeg a lot. I'll try this. 
[23:04:59 CEST] <omerjerk> thanks.
[23:13:33 CEST] <Daemon404> btw what a nice fate sample th intrax8 is...
[23:13:39 CEST] <Daemon404> its a flavoured condom commercial.
[23:18:11 CEST] <durandal_170> omerjerk: what's issue with float support in als?
[23:19:56 CEST] <omerjerk> I used a random wav analyzer software (found on Google), ran it on the output .wav file, and it shows the sampling format as 16-bit int, though it should be 32-bit float. 
[23:20:19 CEST] <omerjerk> I'm properly setting the output format to float already.
[23:20:35 CEST] <omerjerk> I'll try with ffprobe once. 
[23:20:58 CEST] <durandal_170> omerjerk: but you transcode
[23:21:20 CEST] <durandal_170> use -c:a pcm_f32le
[23:21:52 CEST] <omerjerk> ^^ what's this exactly ?
[23:23:10 CEST] <omerjerk> P.S. I'm setting the output format here - https://github.com/omerjerk/FFmpeg/blob/float/libavcodec/alsdec.c#L2143
[23:26:31 CEST] <durandal_170> omerjerk: output codec for wav
[23:31:12 CEST] <cone-749> ffmpeg 03Christophe Gisquet 07master:9630b3fc06d8: x86: lossless audio: SSE4 madd 32bits
[23:54:53 CEST] <kierank> atomnuker: interesting patch
[23:59:11 CEST] <jamrial> kierank: do you have time to fuzz the dca express decoder in the ml?
[23:59:35 CEST] <kierank> Yes I could
[00:00:00 CEST] --- Sun May  8 2016


More information about the Ffmpeg-devel-irc mailing list